square to sine

--- Assuming you want the sine wave to be the same frequency as the square wave's fundamental, run the square wave through a bandpass filter tuned to the square wave's fundamental frequency.

-- John Fields Professional Circuit Designer

The problem with filters is that they work well only over a narrow range of frequencies. Another method that will work over a wider range of frequencies is to run the square wave through two integrators. The first one will convert it to a triangle wave and the second one will convert it to something very close to a sine wave. Of course the amplitude of the resultant sine wave will vary inversely with the frequency. Now you know why function generator ICs put out square waves, triangle waves and sine waves.

"Sandeep" schreef in bericht news: snipped-for-privacy@g49g2000cwa.googlegroups.com...

You just need a good 1kHz filter. Sine will be as stable as your square.

petrus bitbyter

I could be wrong, but wouldn't a low-pass filter be easier to construct, and more to the point?

"Anthony Fremont" wrote

construct,

Never mind I suppose, I just notice a separate thread on SED covering that very topic. It would appear that the bandpass method may be preferred in some situations.

I would use a low pass filter also. It needs to start cutting off somewhat above the fundamental frequency. As the square wave is all the odd harmonics, if your cutoff is below 3 khz (using your 1 khz square wave) then you will have a very good sine wave. after the filter.

Ever since I've heard of the 8038 and the phase-locked loop, I've had a fantasy of locking in a sine wave to some square wave by PLL of some kind.

Set up an 8038 (or maybe these days it's the XR220whatever) as a 1 KHZ oscillator, and run a phase detector output into the freq. control input. Should be a piece of cake. ;-)

Cheers! Rich

On 4 Nov 2005 13:53:34 -0800, via , "skeptic" spake thusly:

OP is only concerned with 1KHz, that's a pretty narrow range.

--
Possibly easier to construct, but its output wouldn\'t really be a
sine wave.

--
The question is, "How good?"

Best help I can offer is to suggest that you disclose the secret operation that you're trying to perform. You're question is incomplete. The solution depends a LOT on unstated requirements.

The simplest thing I can think of is a high Q resonator. I'd make a simple coaxial cavity resonator. But it'd have to be somewhat bent to compensate for the curved surface of the earth. mike

--
Wanted, Serial cable for Dell Axim X5 PDA.
Return address is VALID but some sites block emails
with links.  Delete this sig when replying.
FS 500MHz Tek DSOscilloscope TDS540 Make Offer
Bunch of stuff For Sale and Wanted at the link below.
MAKE THE OBVIOUS CHANGES TO THE LINK
http://www.geocities.com/SiliconValley/Monitor/4710/

It would be a sine wave, if the input was a true square wave with +/- swing, otherwise it would be a true sine wave with some DC offset.

The advantage of the bandpass is simply higher Q, but in theory a lowpass form should be superior since the bandpass is wasting half of its slopes on the low side where they do nothing useful. The trick is to use a lowpass with high Q giving a peak at 1 kHz. The fact that there is a plateau on the lower side is of no consequence, since there isn't anything there anyway.

Whether lowpass or bandpass, at high Q there will be the issue of stability. If the tuning drifts slightly, the output amplitude may change substantially. This is one argument for a more-typical flat passband lowpass, since you can make it less sensitive to drift in components or the input square wave frequency... if you are willing to put more stages into the circuit.

The OP doesn't mention where the original square wave comes from. If it is being generated in a circuit under your control, you can instead generate at a much higher frequency and divide it down to 1 kHz. This opens up two new possibilitues:

1) Use a switched capacitor filter where the high frequency becomes the filter clock, insuring solid stability even at high Q.

or 2), you can make a cute little D/A type of circuit where you sum different amounts of the higher components such that the harmonics cancel. I've seen several designs like this over the years, but I don't know what they would be called to search for them in Google. The big advantage is that they work over a broad range of frequencies, and they virtually elimiante all harmonics below 2^N - 1 where N is the number of divider stages in the original (I think!). Basically, a handful of resistors and a few divider stages is all this takes, and you can follow it with a simple lowpass to remove the higher trash.

Best regards,

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

Home of DaqGen, the FREEWARE signal generator

construct,

How so? I can see that there is a slight difference in the skirt slopes of the two filters. But since no filter is perfect, can you really say that the output of the low pass filter is not a sine wave, but the output of the bandpass filter is? What is the distinguishing factor?

--- My thinking was that with a simple single-pole lowpass with a square-wave input:

SQIN>---[R]--+--->OUT | [C] | GND>---------+--->GND

You wind up with an integrator which merely charges and discharges the cap as the square wave makes its excursions.

Run this through Linear's simulator to see what I mean:

Version 4 SHEET 1 880 680 WIRE -32 192 -32 160 WIRE -32 304 -32 272 WIRE -32 336 -32 304 WIRE 48 160 -32 160 WIRE 160 160 128 160 WIRE 160 192 160 160 WIRE 160 304 -32 304 WIRE 160 304 160 256 FLAG -32 336 0 SYMBOL voltage -32 176 R0 WINDOW 3 24 104 Invisible 0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName V1 SYMATTR Value PULSE(-1 1 0 0 0 5e-4 1e-3) SYMBOL res 144 144 R90 WINDOW 0 -38 58 VBottom 0 WINDOW 3 -36 62 VTop 0 SYMATTR InstName R1 SYMATTR Value 1000 SYMBOL cap 144 192 R0 WINDOW 0 41 34 Left 0 SYMATTR InstName C1 SYMATTR Value 159n TEXT 0 320 Left 0 !.tran 0 .01 0

OTOH, a simple bandpass filter, like this:

SQIN>--[R]--[L]--[C]--+-->OUT | [R] | GND>------------------+-->GND

will give you a beatiful sine wave out:

Version 4 SHEET 1 880 680 WIRE -80 192 -80 160 WIRE -80 288 -80 272 WIRE -80 336 -80 288 WIRE 0 160 -80 160 WIRE 128 160 80 160 WIRE 272 160 208 160 WIRE 368 160 336 160 WIRE 368 192 368 160 WIRE 368 288 -80 288 WIRE 368 288 368 272 FLAG -80 336 0 SYMBOL voltage -80 176 R0 WINDOW 3 24 104 Invisible 0 WINDOW 123 -91 84 Left 0 WINDOW 39 -111 109 Left 0 WINDOW 0 -75 53 Left 0 SYMATTR Value PULSE(-1 1 0 0 0 5e-4 1e-3) SYMATTR Value2 AC 1 SYMATTR InstName V1 SYMBOL cap 272 176 R270 WINDOW 0 32 32 VTop 0 WINDOW 3 0 32 VBottom 0 SYMATTR InstName C1 SYMATTR Value 159n SYMBOL ind 112 176 R270 WINDOW 0 72 56 VTop 0 WINDOW 3 70 55 VBottom 0 SYMATTR InstName L1 SYMATTR Value 159e-3 SYMBOL res 352 176 R0 SYMATTR InstName R2 SYMATTR Value 100 SYMBOL res 96 144 R90 WINDOW 0 -36 59 VBottom 0 WINDOW 3 -31 59 VTop 0 SYMATTR InstName R1 SYMATTR Value 1000 TEXT 48 320 Left 0 !.tran 0 .02 0 TEXT -24 352 Left 0 !;ac oct 128 100 10000

-- John Fields Professional Circuit Designer

--
Please see my reply to Bob Masta\'s post in this ng.

Two op amps in series with 100K on each (-) input and feedback from the output to the (-) input through a .01mF capacitor. Both (+) inputs grounded? Or should the values be changed?

R >

To get the lowpass to work here, the corner frequency needs to be just above the 1 kHz square wave frequency. An RC circuit like the above is only a good integrator approximation if its time constant is long, such that the effective corner is well below the input frequency.

A simple RC is probably not going to be adequate to recover a very good sine wave approximation, though.

Best regards,

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

Home of DaqGen, the FREEWARE signal generator

The OP can see this in action (or a similar scheme in action) by searching for and viewing the ICL8038 datasheet. It uses this scheme by making a triangle wave from a couple of current sources and a flipflop, and runs it through an array of PNP transistors. There is a schematic for the chip in the datasheet.

The ICL8038 is now quite pricy, since it appears to be out of production. However, so you may as well go for the Cadillac:

or

For $20, you can get a signal generator on a chip that goes from .1Hz to

20MHz. It also appears to feature a PLL input, which allows the chip to lock the chip to an external reference, such as a square wave. I haven't spent much time looking at it, but it might be just the thing the OP needs.

--
Regards,
  Bob Monsen

When earlier, new functions were invented, the purpose was to apply them.  
Today, on the contrary, one constructs functions to contradict the 
conclusions of our predecessors and one will never be able to apply them for 
any other purpose.
- Poincare

As has been mentioned, you can filter the higher harmonics out using a band pass or preferably a low pass filter but that only works at a small range of frequencies. Depending on the required purity of the sine wave, filtering may prove to be difficult requiring a steep slope multi-pole filter to clean out harmonics. If the square wave has any asymmetry, even a percent or two, there will be a significant second harmonic present. This is a bitch to get rid of because of its closeness to the fundamental. Even the third harmonic can be troublesome.

Another idea is the synthesizer scheme: Use the squarewave to switch an integrator to ramp positive then negative on each half cycle producing a triangle wave from the square wave. Run the triangle wave into an automatic gain stage to generate a constant amplitude. Use a diode-resistor shaping network to round the triangle wave into a sine wave form. Run this into another variable gain stage controlled by the amplitude of the original square wave to re-constitute the original amplitude of the square wave for the sine wave.

This scheme creates a sine and triangle wave out of the square wave and can work over a range of frequencies. If the amplitude is not important, the variable gain stages can be ommited. Bob